Efficient LLM Inference using Dynamic Input Pruning and Cache-Aware Masking

• URL: http://arxiv.org/abs/2412.01380v2
• Date: Thu, 03 Apr 2025 13:28:51 GMT
• Title: Efficient LLM Inference using Dynamic Input Pruning and Cache-Aware Masking
• Authors: Marco Federici, Davide Belli, Mart van Baalen, Amir Jalalirad, Andrii Skliar, Bence Major, Markus Nagel, Paul Whatmough,
• Abstract summary: We introduce a predictor-free dynamic sparsification approach, which preserves accuracy with minimal fine-tuning.<n>We describe a novel cache-aware masking strategy, which considers the cache state and activation magnitude to further increase cache hit rate.<n>DIP achieves a 46% reduction in memory and 40% increase in throughput with $$ 0.1 loss in perplexity when compared to streaming the dense model from Flash.
• Score: 12.664307714758843
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While mobile devices provide ever more compute power, improvements in DRAM bandwidth are much slower. This is unfortunate for large language model (LLM) token generation, which is heavily memory-bound. Previous work has proposed to leverage natural dynamic activation sparsity in ReLU-activated LLMs to reduce effective DRAM bandwidth per token. However, more recent LLMs use SwiGLU instead of ReLU, which results in little inherent sparsity. While SwiGLU activations can be pruned based on magnitude, the resulting sparsity patterns are difficult to predict, rendering previous approaches ineffective. To circumvent this issue, our work introduces Dynamic Input Pruning (DIP): a predictor-free dynamic sparsification approach, which preserves accuracy with minimal fine-tuning. DIP can further use lightweight LoRA adapters to regain some performance lost during sparsification. Lastly, we describe a novel cache-aware masking strategy, which considers the cache state and activation magnitude to further increase cache hit rate, improving LLM token rate on mobile devices. DIP outperforms other methods in terms of accuracy, memory and throughput trade-offs across simulated hardware settings. On Phi-3-Medium, DIP achieves a 46\% reduction in memory and 40\% increase in throughput with $<$ 0.1 loss in perplexity when compared to streaming the dense model from Flash. The open source code for HW simulator, methods, and experiments in this paper is available at https://github.com/Qualcomm-AI-research/dynamic-sparsity .

Related papers

• 70% Size, 100% Accuracy: Lossless LLM Compression for Efficient GPU Inference via Dynamic-Length Float [71.43026659686679]
  Large Language Models (LLMs) have grown rapidly in size, creating challenges for efficient deployment on resource-constrained hardware. We introduce Dynamic-Length Float (DFloat11), a compression framework that reduces LLM size by 30% while preserving outputs that are bit-for-bit identical to the original model.
  arXiv  Detail & Related papers  (2025-04-15T22:38:38Z)
• Scaling Up On-Device LLMs via Active-Weight Swapping Between DRAM and Flash [21.8915200300114]
  Large language models (LLMs) are increasingly being deployed on mobile devices, but the limited DRAM capacity constrains the deployable model size. This paper introduces ActiveFlow, the first LLM inference framework that can achieve adaptive DRAM usage for modern LLMs.
  arXiv  Detail & Related papers  (2025-04-11T09:26:47Z)
• Dynamic Low-Rank Sparse Adaptation for Large Language Models [54.1231638555233]
  Low-rank Sparse Adaptation (LoSA) is a novel method that seamlessly integrates low-rank adaptation into sparse LLM sparsity. LoSA dynamically sparsifies the LoRA outcomes based on the corresponding sparse weights during fine-tuning. LoSA can efficiently boost the efficacy of sparse LLMs within a few hours, without introducing any additional inferential burden.
  arXiv  Detail & Related papers  (2025-02-20T18:37:32Z)
• Activation Sparsity Opportunities for Compressing General Large Language Models [4.5624217435826]
  This work systematically investigates the tradeoff between enforcing activation sparsity and perplexity (accuracy) on state-of-the-art AI models. Our empirical analysis demonstrates that we can obtain around 50% of main memory and computing reductions for critical FFN components with negligible accuracy degradation.
  arXiv  Detail & Related papers  (2024-12-13T02:26:54Z)
• Sparsing Law: Towards Large Language Models with Greater Activation Sparsity [62.09617609556697]
  Activation sparsity denotes the existence of substantial weakly-contributed elements within activation outputs that can be eliminated. We propose PPL-$p%$ sparsity, a precise and performance-aware activation sparsity metric. We show that ReLU is more efficient as the activation function than SiLU and can leverage more training data to improve activation sparsity.
  arXiv  Detail & Related papers  (2024-11-04T17:59:04Z)
• Enhancing Zeroth-order Fine-tuning for Language Models with Low-rank Structures [21.18741772731095]
  Zeroth-order (ZO) algorithms offer a promising alternative by approximating gradients using finite differences of function values. Existing ZO methods struggle to capture the low-rank gradient structure common in LLM fine-tuning, leading to suboptimal performance. This paper proposes a low-rank ZO algorithm (LOZO) that effectively captures this structure in LLMs.
  arXiv  Detail & Related papers  (2024-10-10T08:10:53Z)
• Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters [20.093224415258174]
  Activation sparsity is determined by activation functions, and commonly used ones like SwiGLU and GeGLU exhibit limited sparsity. We propose a novel dReLU function, which is designed to improve LLM activation sparsity, along with a high-quality training data mixture ratio. On mobile phones, our TurboSparse-Mixtral-47B achieves an inference speed of 11 tokens per second.
  arXiv  Detail & Related papers  (2024-06-10T01:21:59Z)
• Enabling High-Sparsity Foundational Llama Models with Efficient Pretraining and Deployment [56.44025052765861]
  Large language models (LLMs) have revolutionized Natural Language Processing (NLP), but their size creates computational bottlenecks. We introduce a novel approach to create accurate, sparse foundational versions of performant LLMs. We show a total speedup on CPUs for sparse-quantized LLaMA models of up to 8.6x.
  arXiv  Detail & Related papers  (2024-05-06T16:03:32Z)
• FFN-SkipLLM: A Hidden Gem for Autoregressive Decoding with Adaptive Feed Forward Skipping [49.66872823080736]
  Autoregressive Large Language Models (e.g., LLaMa, GPTs) are omnipresent achieving remarkable success in language understanding and generation. To mitigate overload incurred during generation, several early-exit and layer-dropping strategies have been proposed. We propose FFN-SkipLLM, which is an input-adaptive feed-forward skipping strategy.
  arXiv  Detail & Related papers  (2024-04-05T02:35:43Z)
• HiRE: High Recall Approximate Top-$k$ Estimation for Efficient LLM Inference [68.59839755875252]
  HiRE comprises of two novel components: (i) a compression scheme to cheaply predict top-$k$ rows/columns with high recall, followed by full computation restricted to the predicted subset, and (ii) DA-TOP-$k$: an efficient multi-device approximate top-$k$ operator. We demonstrate that on a one billion parameter model, HiRE applied to both the softmax as well as feedforward layers, achieves almost matching pretraining and downstream accuracy, and speeds up inference latency by $1.47times$ on a single TPUv5e device.
  arXiv  Detail & Related papers  (2024-02-14T18:04:36Z)
• ReLU$^2$ Wins: Discovering Efficient Activation Functions for Sparse LLMs [91.31204876440765]
  We introduce a general method that defines neuron activation through neuron output magnitudes and a tailored magnitude threshold. To find the most efficient activation function for sparse computation, we propose a systematic framework. We conduct thorough experiments on LLMs utilizing different activation functions, including ReLU, SwiGLU, ReGLU, and ReLU$2$.
  arXiv  Detail & Related papers  (2024-02-06T08:45:51Z)
• Dynamic Sparse No Training: Training-Free Fine-tuning for Sparse LLMs [67.38165028487242]
  We introduce Dynamic Sparse No Training (DSnoT), a training-free fine-tuning approach to fine-tune large language models (LLMs) Inspired by the Dynamic Sparse Training, DSnoT minimizes the reconstruction error between the dense and sparse LLMs. Our paper offers fresh insights into how to fine-tune sparse LLMs in an efficient training-free manner and open new venues to scale the great potential of sparsity to LLMs.
  arXiv  Detail & Related papers  (2023-10-13T07:38:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.